Daniel Reetz, the founder of the DIY Book Scanner community, has recently started making videos of prototyping and shop tips. If you are tinkering with a book scanner (or any other project) in your home shop, these tips will come in handy. https://www.youtube.com/channel/UCn0gq8 ... g_8K1nfInQ

Best way to get most accurate Split Page step?

Scan Tailor specific announcements, releases, workflows, tips, etc. NO FEATURE REQUESTS IN THIS FORUM, please.
Post Reply
Marcus Eastty
Posts: 9
Joined: 13 Apr 2015, 01:45
Number of books owned: 0
Country: United States

Best way to get most accurate Split Page step?

Post by Marcus Eastty » 29 Apr 2015, 03:14

Hey there!

So I am building my own scanner, planning to use a set-up similar to David Landin's. Book open between two 45º blocks, a 90º platen that lowers into the book, flattening both pages at once. However, this set-up will have one camera (a 13 MP smartphone) attached to a crossbar rather than having two camera support bars.

Because of this, the photos will come out looking something like this quickly thrown together example...
So this set-up will depend on heavier use of post-processing than the dual-camera set-ups. The plan right now is to...
  • Run it through ScanTailor to split the pages. (This will also order the pages more simply than the simple, single-camera, all-odds-then-all-evens setup I had used in my first project.)
  • Take the output files, run them through YASW to batch dekeystone and crop the images. If my set-up works as I envision, the relationship between the phone and platen will be constant, so these steps will *hopefully* be quite accurate in batch.
  • Then, once there's a nice, rectangular image, send them through ScanTailor (again) for more accurate content recognition.
The first step, using the Split Pages function in ScanTailor, has not been as consistent as I have been hoping. I don't know how the software recognizes sides to the page, but I am curious... the platen will be secured at 90º by hardware, but if I run some caulk or silicone (or whatever would best be used for non-glare styrene) down the length of the connection of the platen, might ScanTailor better recognize the devide in the pages? This could make scanning some books with a small margin difficult, but I am envisioning a very thin strip, such as the following.
Use p.jpg
My experience with ScanTailor is very limited at this point, so I hope somebody with experience or knowledge as to how ScanTailor splits pages could chime in.

I am very grateful for all of the advice the community has brought to this point, and any further advice will certainly be appreciated!


Posts: 383
Joined: 01 Jun 2014, 17:04
Number of books owned: 1000
Country: United States of America

Re: Best way to get most accurate Split Page step?

Post by duerig » 29 Apr 2015, 11:11

You might get some better results that way. ScanTailor uses Howe line detectiong, so if there is a well-defined line it should find it. Your only trouble after this will be on pages with vertical lines printed on them. If there is more than one vertical line, then it can easily pick the wrong one.

Another possible solution if your camera is fixed relative to the platen might be to try to do a mass crop of the book. Find the line once and then run a batch image manipulation program (before running ST) to crop along that line to get all of the pages. But this doesn't work if the platen moves a little from one image to another.


Post Reply